Skip to content

SimpleProducer randomization of initial round robin ordering #139

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
May 7, 2014
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 15 additions & 1 deletion kafka/producer.py
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@

import logging
import time
import random

from Queue import Empty
from collections import defaultdict
Expand Down Expand Up @@ -180,14 +181,20 @@ class SimpleProducer(Producer):
batch_send - If True, messages are send in batches
batch_send_every_n - If set, messages are send in batches of this size
batch_send_every_t - If set, messages are send after this timeout
random_start - If true, randomize the initial partition which the
the first message block will be published to, otherwise
if false, the first message block will always publish
to partition 0 before cycling through each partition
"""
def __init__(self, client, async=False,
req_acks=Producer.ACK_AFTER_LOCAL_WRITE,
ack_timeout=Producer.DEFAULT_ACK_TIMEOUT,
batch_send=False,
batch_send_every_n=BATCH_SEND_MSG_COUNT,
batch_send_every_t=BATCH_SEND_DEFAULT_INTERVAL):
batch_send_every_t=BATCH_SEND_DEFAULT_INTERVAL,
random_start=False):
self.partition_cycles = {}
self.random_start = random_start
super(SimpleProducer, self).__init__(client, async, req_acks,
ack_timeout, batch_send,
batch_send_every_n,
Expand All @@ -198,6 +205,13 @@ def _next_partition(self, topic):
if topic not in self.client.topic_partitions:
self.client.load_metadata_for_topics(topic)
self.partition_cycles[topic] = cycle(self.client.topic_partitions[topic])

# Randomize the initial partition that is returned
if self.random_start:
num_partitions = len(self.client.topic_partitions[topic])
for _ in xrange(random.randint(0, num_partitions-1)):
self.partition_cycles[topic].next()

return self.partition_cycles[topic].next()

def send_messages(self, topic, *msg):
Expand Down
81 changes: 58 additions & 23 deletions test/test_integration.py
Original file line number Diff line number Diff line change
Expand Up @@ -287,21 +287,26 @@ def test_simple_producer(self):
producer = SimpleProducer(self.client)
resp = producer.send_messages(self.topic, "one", "two")

# Will go to partition 0
partition_for_first_batch = resp[0].partition

self.assertEquals(len(resp), 1)
self.assertEquals(resp[0].error, 0)
self.assertEquals(resp[0].offset, 0) # offset of first msg

# Will go to partition 1
# ensure this partition is different from the first partition
resp = producer.send_messages(self.topic, "three")
partition_for_second_batch = resp[0].partition
self.assertNotEquals(partition_for_first_batch, partition_for_second_batch)

self.assertEquals(len(resp), 1)
self.assertEquals(resp[0].error, 0)
self.assertEquals(resp[0].offset, 0) # offset of first msg

fetch1 = FetchRequest(self.topic, 0, 0, 1024)
fetch2 = FetchRequest(self.topic, 1, 0, 1024)
fetch_resp1, fetch_resp2 = self.client.send_fetch_request([fetch1,
fetch2])
fetch_requests = (
FetchRequest(self.topic, partition_for_first_batch, 0, 1024),
FetchRequest(self.topic, partition_for_second_batch, 0, 1024),
)
fetch_resp1, fetch_resp2 = self.client.send_fetch_request(fetch_requests)
self.assertEquals(fetch_resp1.error, 0)
self.assertEquals(fetch_resp1.highwaterMark, 2)
messages = list(fetch_resp1.messages)
Expand All @@ -314,11 +319,12 @@ def test_simple_producer(self):
self.assertEquals(len(messages), 1)
self.assertEquals(messages[0].message.value, "three")

# Will go to partition 0
# Will go to same partition as first batch
resp = producer.send_messages(self.topic, "four", "five")
self.assertEquals(len(resp), 1)
self.assertEquals(resp[0].error, 0)
self.assertEquals(resp[0].offset, 2) # offset of first msg
self.assertEquals(resp[0].partition, partition_for_first_batch)

producer.stop()

Expand Down Expand Up @@ -396,14 +402,25 @@ def test_acks_none(self):
resp = producer.send_messages(self.topic, "one")
self.assertEquals(len(resp), 0)

fetch = FetchRequest(self.topic, 0, 0, 1024)
fetch_resp = self.client.send_fetch_request([fetch])
# fetch from both partitions
fetch_requests = (
FetchRequest(self.topic, 0, 0, 1024),
FetchRequest(self.topic, 1, 0, 1024),
)
fetch_resps = self.client.send_fetch_request(fetch_requests)

self.assertEquals(fetch_resp[0].error, 0)
self.assertEquals(fetch_resp[0].highwaterMark, 1)
self.assertEquals(fetch_resp[0].partition, 0)
# determine which partition was selected (due to random round-robin)
published_to_resp = max(fetch_resps, key=lambda x: x.highwaterMark)
not_published_to_resp = min(fetch_resps, key=lambda x: x.highwaterMark)
self.assertNotEquals(published_to_resp.partition, not_published_to_resp.partition)

messages = list(fetch_resp[0].messages)
self.assertEquals(published_to_resp.error, 0)
self.assertEquals(published_to_resp.highwaterMark, 1)

self.assertEquals(not_published_to_resp.error, 0)
self.assertEquals(not_published_to_resp.highwaterMark, 0)

messages = list(published_to_resp.messages)
self.assertEquals(len(messages), 1)
self.assertEquals(messages[0].message.value, "one")

Expand All @@ -415,12 +432,14 @@ def test_acks_local_write(self):
resp = producer.send_messages(self.topic, "one")
self.assertEquals(len(resp), 1)

fetch = FetchRequest(self.topic, 0, 0, 1024)
partition = resp[0].partition

fetch = FetchRequest(self.topic, partition, 0, 1024)
fetch_resp = self.client.send_fetch_request([fetch])

self.assertEquals(fetch_resp[0].error, 0)
self.assertEquals(fetch_resp[0].highwaterMark, 1)
self.assertEquals(fetch_resp[0].partition, 0)
self.assertEquals(fetch_resp[0].partition, partition)

messages = list(fetch_resp[0].messages)
self.assertEquals(len(messages), 1)
Expand All @@ -435,12 +454,14 @@ def test_acks_cluster_commit(self):
resp = producer.send_messages(self.topic, "one")
self.assertEquals(len(resp), 1)

fetch = FetchRequest(self.topic, 0, 0, 1024)
partition = resp[0].partition

fetch = FetchRequest(self.topic, partition, 0, 1024)
fetch_resp = self.client.send_fetch_request([fetch])

self.assertEquals(fetch_resp[0].error, 0)
self.assertEquals(fetch_resp[0].highwaterMark, 1)
self.assertEquals(fetch_resp[0].partition, 0)
self.assertEquals(fetch_resp[0].partition, partition)

messages = list(fetch_resp[0].messages)
self.assertEquals(len(messages), 1)
Expand All @@ -456,17 +477,31 @@ def test_async_simple_producer(self):
# Give it some time
time.sleep(2)

fetch = FetchRequest(self.topic, 0, 0, 1024)
fetch_resp = self.client.send_fetch_request([fetch])
# fetch from both partitions
fetch_requests = (
FetchRequest(self.topic, 0, 0, 1024),
FetchRequest(self.topic, 1, 0, 1024),
)
fetch_resps = self.client.send_fetch_request(fetch_requests)

self.assertEquals(fetch_resp[0].error, 0)
self.assertEquals(fetch_resp[0].highwaterMark, 1)
self.assertEquals(fetch_resp[0].partition, 0)
# determine which partition was selected (due to random round-robin)
published_to_resp = max(fetch_resps, key=lambda x: x.highwaterMark)
not_published_to_resp = min(fetch_resps, key=lambda x: x.highwaterMark)
self.assertNotEquals(published_to_resp.partition, not_published_to_resp.partition)

messages = list(fetch_resp[0].messages)
self.assertEquals(published_to_resp.error, 0)
self.assertEquals(published_to_resp.highwaterMark, 1)

self.assertEquals(not_published_to_resp.error, 0)
self.assertEquals(not_published_to_resp.highwaterMark, 0)

messages = list(published_to_resp.messages)
self.assertEquals(len(messages), 1)
self.assertEquals(messages[0].message.value, "one")

messages = list(not_published_to_resp.messages)
self.assertEquals(len(messages), 0)

producer.stop()

def test_async_keyed_producer(self):
Expand Down